Huge memory leak using docker image 1.4.1 from ghcr

[MaitreGEEK]

Describe the bug
Huge memory leak on the command using docker image 1.4.1
It's in a proxmox and the gpu is passed through. There's 16GB RAM and 14 are used by the nvidia-smi command used by the docker image

Expected behavior
It shouldn't use that much memory

Console output

Model and Version

• GPU Model GeForce RTX 2060 Super
• App version and architecture [e.g. v1.4.1 - linux_x86_64]
• Installation method ghcr.io image
• OS Debian 13 VM under Proxmox v9.0.11
• Nvidia GPU driver version Linux driver nvidia-driver-550.163.01

Additional context

[utkuozdemir]

Hi, I made a PR which enabled pprof, #375.
And made a local build of it: ghcr.io/utkuozdemir/nvidia_gpu_exporter:1.4.1-SNAPSHOT-328d96f-amd64

Can you give it a try? You can build it yourself if you want. When you observe high memory usage, if you can get a heap dump (from /debug/pprof/heap path) and provide it here, that would be super helpful.

[MaitreGEEK]

Okay, going to send it

[MaitreGEEK]

It seems that with this image ghcr.io/utkuozdemir/nvidia_gpu_exporter:1.4.1-SNAPSHOT-328d96f-amd64 I have no issue with memory consumption

[utkuozdemir]

I guess then it happens only sometimes - if you see the PR, there's no real change in it.
I'm gonna merge that and make a release. You can upgrade to that. When the issue happens next time, pprof endpoint will be available, so you can capture a heap dump and inspect/create an issue with it.

[MaitreGEEK]

Alright, thanks